When will the service be available? Tweet

[renderingz]

Well, that's a pretty good presentation..... <Question re: 15min frames X 1000 @250 Octanebench>....

<part of the answer>....cost you $750, and at roughly about two three days render time, again on $10 an hour. I guess you could book two 800 octane bench buckets for $20 an hour and get it done half the time...

With all due respect I think your calculations/conclusions are totally missing the (main) point (or strength) or ORC:

Any render job (single still or complete set of frames) which takes over an hour you would pay the higher rate for MORE buckets...

The example above is 78.125 hours by my calculations (Renderingz, you rounded this to 75 hours)
As has been noted; animations is the real strength of ORC because you easily get over the 1 hour 'job' size.
In this example you would actually pay for 79 buckets; at a cost of $790/hour, and finishing the job in ONE HOUR...

You would never pay $790 for 79 hours (x1 bucket) and sit around waiting...


To fine tune the example;
If you had 400 frames which only take 3 minutes on a OB 250 system;
375min on Single ORC bucket (or 6.25 bucket hours)

So you buy 7 buckets for an hour; finish the job in 53.5 minutes...
Your "waste" is:
6.25Bhr req / 7.0Bhr booked = 0.89 usage * 60 (minutes bought) = 53.5mins.. ie 6.5mins your buckets of GPU's sit there idle...(10%)

FYI: waste for the original 78.125hr example is:
78.125/79=98.89% or 1.11%


I could spreadsheet this, because there would be a sweet spot for waste on every job... I assume the Otoy guys have some pre-job estimator that gives the artist instant feedback/graph so thy can work out the Time/waste ratio they are comfortable with?

Yeah I get that, I imagine there must be a premium for booking up more buckets though - I guess that accounts to a priority service, otherwise it would cost you the same to have it done in an hour as it would 79 - would be good to hear from the team with regards to that.

I think I read or heard in the presentation they're not charging for CPU time and scene export..? Is that right?

[xcaseyx]

Im more curious about the costs for enterprise licensing and how much it will cost for us to make our own 'orc'

the new p100 chips from nvidia will most certainly make their way to AWS. with THAT being said, it would be a lot cheaper to fold your own ec2 cluster and use deadline when that does happen.

Currently if you wanted to network 12 g2.2xlarge gpus on aws together, (octane bench of 428) it would cost you ~3$ an hour (4 if you want to make sure you're not getting booted)

Amazons EC2 gpu instances are CRAZY out of date... With what just went down at GTC, those p100s aren't that far off. (which will make these g2.xXlarge clusters EVEN cheaper.

so... what does building your own ORC look like cost wise?

[Goldorak]

Is the data that is send to the Cloud fully encripted so that the clients are less concerned in terms of NDA breaches?

Yes - to MPAA specs. For safety, any Octane scene assets or data you send to ORC are not exposed or accessible for re-download - you can just delete it or re-render a scene. This means even if your ORC login is compromised your data is not exposed.

[Refracty]

That is really good news.

Is the data that is send to the Cloud fully encripted so that the clients are less concerned in terms of NDA breaches?

Yes - to MPAA specs. For safety, any Octane scene assets or data you send to ORC are not exposed or accessible for re-download - you can just delete it or re-render a scene. This means even if your ORC login is compromised your data is not exposed.